MURDOCH RESEARCH REPOSITORY

This is the author’s final version of the work, as accepted for publication following peer review but without the publisher’s layout or pagination.

The definitive version is available at http://dx.doi.org/10.1016/j.apcata.2015.09.039

Sánchez, G., Dlugogorski, B.Z., Kennedy, E.M. and Stockenhuber, M. (2016) Zeolite-supported iron catalysts for allyl alcohol

synthesis from glycerol. Applied Catalysis A: General, 509. pp. 130-142.

http://researchrepository.murdoch.edu.au/29136/

Accepted Manuscript

Title: Zeolite-supported iron catalysts for allyl alcoholsynthesis from glycerol

Author: G. Sanchez B.Z. Dlugogorski E.M. Kennedy M.Stockenhuber

PII: S0926-860X(15)30172-1DOI: http://dx.doi.org/doi:10.1016/j.apcata.2015.09.039Reference: APCATA 15571

To appear in: Applied Catalysis A: General

Received date: 17-6-2015Revised date: 23-9-2015Accepted date: 24-9-2015

Please cite this article as:G.Sanchez,B.Z.Dlugogorski, E.M.Kennedy,M.Stockenhuber,Zeolite-supported iron catalysts for allyl alcohol synthesis from glycerol, AppliedCatalysis A, General http://dx.doi.org/10.1016/j.apcata.2015.09.039

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

1

Zeolite-supported iron catalysts for allyl alcohol synthesis

from glycerol

G. Sancheza, B. Z. Dlugogorskib, E.M. Kennedya, M. Stockenhubera*

a Priority Research Centre for Energy (PRCfE), The University of Newcastle, Callaghan, NSW 2308, Australia b School of Engineering and Information Technology, Murdoch University, Murdoch, WA 6150, Australia. *Corresponding author: michael.stockenhuber@newcastle.edu.au

Dochead : Research paper

Highlights

Increased iron loading on MFI improved allyl alcohol yield;

Alkaline preparation conditions engendered the zeolite with mesoporosity;

Rubidium deposition enhanced allyl alcohol yield and decreased acrolein yield;

Alkali post synthesis modification reduced the concentration of acid sites.

Abstract

Under most reaction conditions studied, acrolein is reported as the primary product in the

conversion of glycerol over zeolites. In such processes, acrolein forms at relatively high yields,

with negligible allyl alcohol selectivity. In this contribution, we report the development of

ZSM5-supported iron catalysts, modified by rubidium deposition, as stable materials for

production of allyl alcohol from glycerol. Our results demonstrate a reduced rate of formation

of acrolein over modified catalysts. Both unmodified and modified catalysts were analysed by

inductively coupled plasma optical emission spectrometry, nitrogen adsorption, scanning

electron microscope, X-ray diffraction, ammonia temperature programmed desorption, X-ray

photoelectron spectroscopy and ultraviolet-visible spectroscopy. These techniques revealed

that differences in product distribution and catalyst performance are due to the combined

effects of iron loading, catalyst acidity and changes in the porosity of the catalyst.

2

Keywords: Allyl Alcohol; glycerol conversion; transition metal zeolites;

acrolein; acidity; basicity

Graphical abstract

3

1. Introduction

The size and shape of zeolites pores are known to influence the product distribution in reactions

such as the conversion of glucose to aromatics or the synthesis of glycerol monolaurate from

glycerol and lauric acid [1, 2]. For “bulky” molecules, zeolitic channels control the rate of

internal diffusion of reactants within the structure (based on steric effects), and limit product

formation, usually as a function of their effective kinetic diameters. Pore size and shape are

also known to favour certain transition states affecting, consequently, final product distribution

[2]. In the conversion of glycerol to allyl alcohol, acrolein is one of the primary (and usually

undesirable) by-products. When considering the influence of shape selectivity on glycerol

conversion, a cursory analysis might suggest these factors would be irrelevant for species such

as allyl alcohol and acrolein given that, in addition to being molecules of similar composition

and conformation, they both have small molecular dimensions. However when converting

glycerol over a variety of catalysts, oligomers of glycerol [3] as well as cyclic compounds and

dimers are present as products from the reaction of glycerol, glycerol/acrolein and

glycerol/acetaldehyde [4]. The presence of such compounds positions zeolites as potential

catalysts for allyl alcohol production from glycerol, since the presence of well defined

micropores can reduce (due to shape selectivity) the formation of those “bulky” molecules

which appear to be involved with, and indeed facilitate, catalyst deactivation. Thus, zeolites

have a major potential advantage over other commonly used supports (most notably alumina

[5] and zirconia [6]). To the best of our knowledge, investigations aimed at the use of zeolites

in the conversion of glycerol to allyl alcohol have not been reported.

4

The acid and base properties of catalysts are important features to consider in the conversion

of glycerol to allyl alcohol. Catalyst acidity has been correlated with acrolein formation, while

basicity was shown to drive the reaction towards allyl alcohol formation [5, 7]. Zeolite acidity

can be typically reduced by dealumination [8, 9], modification with phosphorus [10], alkali

treatment [11], alkali metal exchange [12] amongst other methods. In the present research, we

have modified the acidity of an iron-exchanged zeolite via rubidium deposition, aiming to

enhance the rate of formation of allyl alcohol from glycerol.

In most applications of zeolites to glycerol conversion, dehydration is found to occur.

Consequently, a high selectivity towards acrolein is observed, whereas the rate of formation of

allyl alcohol is negligible. The presence of C=C and C-OH groups in allyl alcohol promotes

the occurrence of additions, substitutions, decompositions, oxidations, re-arrangements and

polymerisation reactions which make of this commodity an important chemical intermediate

[13]. Allyl alcohol derivatives are present in perfume, pharmaceutical and food formulations.

In the chemical industry, allyl alcohol is required for the production of allyl diglycol carbonate,

allyl glycidyl ether, allyl methacrylate, diallyl phthalate, 1,4-butanediol, 2-methyl-1,3-

propanediol, poly(styrene-allyl alcohol), etc., which are used as lens, coupling agents,

plasticisers, crosslinking agents and coating additives [14, 15]. This broad range of

applications creates a market requirement for allyl alcohol which has a higher commercial

value than other value-added glycerol products such as acrolein [16].

Current industrial processes for the production of allyl alcohol are fossil fuel-based routes that

involve converting propylene. The oxidation of propylene to acrolein, followed by subsequent

reduction is one example. A second approach involves propylene oxide isomerisation over a

lithium phosphate catalyst at 300 °C with subsequent purification. Moreover, when allyl

chloride undergoes alkaline hydrolysis, allyl alcohol is synthesised in yields of 92 – 93 %.

Likewise, allyl alcohol can also be produced by further processing the liquid effluents of the

5

palladium catalysed acetoxylation of propylene [17]. Propylene is conventionally obtained

from refining processes or by cracking of heavy liquids and its supply has historically satisfied

the global market. However, recently in the US, the production of propylene has decreased due

to the wide use of shale gas and the simultaneous production of ethane, propane and butane

that are cracked instead of heavy hydrocarbons. This together with liquid petroleum gas (LPG)

cracking in Europe has promoted an increase in global propylene prices [18] since propylene

is highly demanded in the manufacture of several chemicals [19]. Considering that

conventional methods for allyl alcohol production use fossil fuel-based propylene as feedstock

and that besides propylene high prices, environmental and fuel security aspects are important

limitations of these processes, investigating bio-routes for the synthesis of allyl alcohol have

recently gained significant interest in the industrial context. In a continuous flow system, the

active phases for the selective formation of allyl alcohol previously reported are iron [5, 7, 20]

and MoO3 - WO3 [21]. In this contribution, we report, for the first time, the synthesis and

modification (in terms of mesoporosity and acidity) of stable exchanged iron zeolites for direct

conversion of glycerol to allyl alcohol in a flow reactor configuration under relatively mild

reaction conditions.

2. Experimental

2.1 Catalyst preparation

Zeolites (NH4-ZSM5 SiO2/Al2O3 = 30) were obtained from Zeolyst and calcined for 5 hours in

air flow at a rate of 5 °C min-1 in order to generate the proton form of the zeolite. Iron was

deposited on the zeolite using various methods. Approximately 10 g of the support and 3.97 g

of 98 % Fe(NO3)3·9H2O (Sigma Aldrich) crystals were mixed and then milled using a mortar

and pestle as part of a standard solid state ion exchange (SSE). A second catalyst was

synthetised by non-aqueous ion exchange (non-aqueous IE), where 3.97 g of the iron salt was

6

dissolved in methanol and the solution dried using CaSO4 (previously dehydrated at 220 °C for

two hours). Following vacuum filtration, 10 g of HZSM-5 was added to the filtrate and

continuously stirred overnight. The third catalyst was prepared adopting an alkaline ion

exchange method described elsewhere [22]. Approximately 10 g of HZSM-5 was added to an

aqueous Fe(NO3)3·9H2O solution (3.97 g of the salt in 100 g of water) and stirred at 50 °C for

12 h. The pH was controlled (pH = 8) by dropwise addition of ammonium hydroxide (28 – 30

wt % NH3 basis, Sigma Aldrich) as previously reported [23]. The mixture was filtered and the

exchange repeated three times. Following preparation, the catalysts were heated at a rate of 5

°C min-1 to 550 °C and calcined for 5 hours in air.

The calcined alkaline ion exchanged catalyst underwent additional treatment using a solution

of 0.147 g of rubidium nitrate in 100 mL of water. The treatment consisted in washing the

catalyst with the alkali solution, followed by filtration. The filter cake was then dried and

calcined on air. All catalysts were sieved between 250 and 600 µm following calcination.

2.2 Catalyst characterisation

The Fe, Si and Al content of the catalysts as prepared was determined using a Varian Radial

715-ES Inductively Coupled Plasma – Optical Emission Spectrometer with a Sturman-Masters

spray chamber, quartz torch and an SPS 3 autosampler. Sample digestion took place in a

Milestone Start D microwave digestion system with acids in concentrations detailed elsewhere

[24], and thulium was used as internal standard. Surface area measurements were carried out

in a Micromeritics Gemini surface area analyser at -196 °C. A degassing step using a

Micromeritics Vac Prep 061 sample preparation device proceeded nitrogen adsorption at liquid

nitrogen temperature. The t-plot and Barrett-Joyner-Halenda (BJH) models were used to

estimate micro and mesopore volume, as well as pore size distribution. SEM images were

7

captured using a Zeiss Sigma VP FESEM with a secondary detector (SE). Simultaneously,

elemental analysis was conducted with a Bruker light element SSD Energy-dispersive X-ray

spectroscopy (EDS) detector. The crystal structure of fresh and used catalyst was determined

by means of a Phillips X’pert Pro diffractometer using Cu Kα radiation. Measurements were

conducted in the 2θ range of 5° to 109°. Catalyst acidity was studied by temperature

programmed desorption (TPD) of ammonia, in an apparatus described elsewhere [25].

Activation took place at 550 °C under high vacuum which was followed by ammonia

adsorption at 150 °C. Desorption data was collected between 30 °C and 550 °C at a heating

rate of 5 °C min-1 where the ion current used for analysis was m/z = 16. Catalysts were analysed

through X-ray photoelectron spectroscopy (XPS) in a Thermo Scientific ESCALAB250Xi

instrument with a spot size of 500 μm using a mono-chromated Al K alpha source (energy

1486.68 eV). UV-visible spectra were recorded by means of a Cary 6000i UV-Vis-NIR

spectrophotometer (Varian) coupled with a Praying Mantis accessory. Kubelka-Munk function

conversions were applied to all the measurements. Total organic carbon determination in used

catalysts samples was conducted in a LECO TruSpec CN Analyser.

2.3 Catalytic measurements

For each experiment, 5 g of catalyst was charged into the reactor equipped with a catalyst

support (stainless steel) and a fritted disk (quartz). Dimensions are reported in previously

published work [5]. Experiments were conducted at atmospheric pressure and 340 °C in

continuous operation with nitrogen as carrier gas and a constant supply of a 35 wt % glycerol

aqueous solution. Details on mass flow controllers, furnaces and pumps are described

elsewhere [26]. Sampling took place at hourly intervals, and product analyses were off-line.

Liquid samples were collected using a syringe. Methanol was used as solvent for GC-MS

analyses, performed in an Agilent 6890 series GC with an Agilent 5973N detector and a Restek

8

Rtx-200 MS column: 30 m × 0.25 mm ID × 0.5 µm. Helium was used as carrier gas with a

sample injection (1 l) split ratio of 10:1 applied for all analyses. Injector and detector were

maintained at 220 °C and 285 °C respectively, while the oven initial temperature (45 °C) was

held for 5 min, increasing to 115 °C in 15 min and then ramping up to 285 °C at a rate of 10

°C · min-1. Filament and detector were turned off during the elution of the injection solvent (i.e.

methanol). For quantification, cyclohexanone was used as internal standard and GC-FID

analyses were conducted in a 5890A model GC, fitted with a Restek Stabilwax column: 30 m

× 0.32 mm ID × 1 µm, using air, hydrogen and helium with a split ratio of 100:1. Injector and

detector were kept at 300 °C and 320 °C respectively, while an initial temperature of 35 °C

was held for 5 min, ramping to 200 °C at a rate of 10 °C · min-1, holding at this temperature for

20 min. Gas samples were collected in a gas bag and analysed using a Varian 490-GC micro

gas chromatograph and an IR Prestige 21 Shimadzu FTIR QP 5000 apparatus. IR spectra were

processed using QASoft software.

3. Results and discussion

3.1 Catalyst characterisation

The various methods employed to synthesise iron zeolites for glycerol conversion were

found to have a significant influence on the eventual iron loading on the zeolite support, as

determined by inductively coupled plasma optical emission spectrometry (ICP-OES). Catalysts

of low iron content (significantly below the theoretical exchange capacity of H-ZSM5 with

silica to alumina ratio equal to 30) were obtained by ion exchange in the methanol procedure

as evident from Table 1. Higher iron containing materials were prepared by solid state

exchange. Interestingly, using the alkaline solution (NH4OH in water, pH 8) as a solvent instead

of methanol to conduct the ion exchange caused a three-fold increase in the mass of iron present

9

in the resulting iron-loaded zeolite with respect to the latter catalyst. Iwamoto et al. found that

controlling the pH of the solution with NH4OH, KOH, NaOH, Ca(OH)2, Mg(OH)2, Ba(OH)2

or pyridine resulted in the preparation of over-exchanged zeolites [27]. More recently, Tsikoza

et al. synthesised Cu-ZSM5 catalysts, with copper content that exceeded the ion exchange

capacity of the zeolite, in aqueous ammonia solutions of a copper salt with pH of approximately

10 [28]. Repeated ion exchange processes have been associated as well to highly exchanged

materials [29]. In these preparation methods, the zeolite undergoes ion exchange several times

(the catalyst is filtrated following the first ion exchange treatment and the filter cake is placed

again in a fresh solution to repeat the exchange) [29]. While an exchange between metal ions

and countercations on aluminium sites have been reported to occur at low pH, electrostatic

adsorption is suggested to take place at high pH on silanol groups when tetra ammine

complexes are formed reported for platinum and cooper [30]. Under our conditions Fe(OH)3 is

likely to form as opposed to the ammine complex, for which such adsorption on silanol groups

was not thought to occur.

The alkaline medium in which the ion exchange was conducted promotes partial dissolution of

both framework silicon and aluminium, as measured by ICP-OES which revealed a

composition of 30.2 wt % silicon and 1.8 wt % aluminium for the alkaline, iron-exchanged

catalysts. Similar observations have been reported when desilicating various MFI zeolites with

NaOH [31]. Melian et al. desilicated ZSM5 with NaOH to prepare fully exchanged Fe-ZSM5

[32]. When the Si/Al ratio decreases following catalyst treatment as reported elsewhere [32],

an increase in the exchange capacity of the material is observed. For the current experiments

this is not applicable due to simultaneous reduction of the aluminium content.

The degree of the exchange (given by the molar ratios H/Al and Fe/Al) has been calculated

adopting a procedure described elsewhere [33]. Results from ammonia TPD experiments

(Figure S1 ESI) were employed to quantify the remaining protons in the catalyst following iron

10

exchange. It was assumed that high temperature peaks are a measure of the number of residual

protons as previously reported [34]. When considering trivalent cations to be exchanged with

H-ZSM5 SiO2/Al2O3 = 30, the maximum cation to aluminium molar ratio for a full and

stoichiometric exchange is equal to 0.33 according to the following reaction,

퐹푒 + 3[퐻 · 퐴푙푂 ] ⇌ 3퐻 + 퐹푒 · 3퐴푙푂

The Fe/Al molar ratio calculated (using ICP-OES data) for the higher iron content exchanged

ZSM5 was 3.66. However, its H/Al molar ratio was equal to 0.43 which suggests that not all

Fe3+ cations are truly exchanged and are present as ions or precipitated species. It is worthwhile

to mention that dissolution of framework aluminium due to the conditions of the ion exchange

have significant influence in the calculated molar ratios.

Table 1: Composition and textural properties of fresh catalysts

Catalyst

Fe content

(wt %)

Langmuir

Surface area

(m2/g)

Micropore

volume

Mesopore

volume

H-ZSM5 0 473.7 0.14 0.04

ZSM5/Fe non-aqueous IE 0.9 487.9 0.13 0.06

ZSM5/Fe SSE 5.4 435.8 0.12 0.05

ZSM5/Fe alkaline IE 13.4 432.7 0.09 0.17

In general, depositing iron on H-ZSM5 reduces its surface area. However, correlating this with

iron loading is not straightforward, as exceeding the exchange capacity of the zeolite has little

influence on the surface area of the resultant material and is comparable with the surface area

11

of the catalyst prepared by solid state exchange (Table 1). As evident by the high nitrogen

adsorption at low P/Po values, ion exchange is the only method of iron deposition that does

not decrease the microporosity of the starting H-ZSM5 (Error! Reference source not found.).

This was attributable to both low iron content (0.9 wt %) in the prepared catalyst and the nature

of the preparation method, where iron deposits preferentially on ion-exchangeable sites of the

support (equal in number to the aluminium atoms present in the framework) [35]. For the other

catalysts, iron incorporation was thought to cause substantial blocking in micropores, while for

the alkaline ion exchanged catalyst, desilication is invoked to explain the reduction in

microporosity, as observed previously by other authors [31].

. Nitrogen adsorption isotherms for fresh catalysts at -196 °C

The iron alkaline ion exchange procedure not only affected the micropores but induces

mesoporosity (in the range of pores sizes between 50 and 200 Å) into the zeolite, as disclosed

in Figure 1. Framework silicon extraction due to the alkaline conditions imposed during

exchange is a known strategy for generating mesoporosity [36].

60

70

80

90

100

110

120

130

140

0 0.2 0.4 0.6 0.8 1

Qua

ntity

ads

orbe

d (c

m3 /g

)

P/P0

H-ZSM5Fe-ZSM5 1% non-aqueous IEFe-ZSM5 5% SSEFE-ZSM5 13% alkaline IE

12

Figure 1. BJH adsorption pore distribution for fresh catalysts

Figure 3 discloses scanning electron micrographs of the starting material and the synthesised

iron catalysts. Images were collected using a secondary electron detector which explains the

absence of brighter areas corresponding to iron. However, the presence of iron under SEM was

confirmed by Energy Dispersive Spectroscopy (EDS) (Figure 3) which was conducted

following secondary electron imaging for each sample. EDS results were consistent with the

catalyst iron content as determined by ICP. While the morphology of ZSM5-supported iron

catalysts prepared by solid state exchange and non-aqueous ion exchange resemble that of H-

ZSM5, the alkaline treatment procedure resulted in additional catalyst alterations such as de-

agglomeration and a “sponge-like” morphology as reported by Possato et al. for desilicated

zeolites [31]. This type morphology is generally associated with silicon dissolution in the

parent material.

0

0.05

0.1

0.15

0.2

0.25

1 10 100

DV

/log(

D)

Pore size (nm)

ZSM5/Fe 1% non-aqueous IE

ZSM5/Fe 5% SSE

ZSM5/Fe 13% alkaline IE

13

Fresh catalysts as prepared were analysed by x-ray diffraction (Figure 3). Even though phase

changes were not observed where iron was present, the data suggests a reduction of the

crystallinity of H-ZSM5 due to iron deposition through the different exchange methods, which

is in agreement with previous research [31, 37]. The intensity of the reflections are inversely

proportional to the iron concentration of the material, as evident in Figure 3, whereas the

intensity of the background is directly proportional to the iron content, predominately at high

angles (Figure 4). A significant reduction in the intensity of the reflections associated to H-

ZSM5 was observed on samples with high iron loading, which is not only attributed to a higher

iron content but also to the severity of the conditions used during alkaline impregnations, as

reported by other authors [31].

Figure 2. a. SEM-EDS of H-ZSM5 catalyst. b. SEM-EDS of 1 wt % Fe catalyst prepared by a non-aqueous ion exchange. c. SEM-EDS of 5 % Fe-ZSM5 catalyst prepared by solid state exchange. d. SEM-EDS of 13 % Fe-ZSM5 catalyst prepared by alkaline ion exchange

a

200nm

b

c d

14

Figure 3. XRD patterns at low angles of fresh H-ZSM5 and ZSM5/Fe patterns prepared by

different methods.

Figure 4. XRD patterns at high angles of fresh H-ZSM5 and ZSM5/Fe catalysts prepared by different methods.

0

5000

10000

15000

20000

25000

30000

35000

40000

5 15 26 36

Inte

nsity

(a.u

.)

2 Theta (°)

H-ZSM5 fresh

ZSM5/Fe 1% non-aqueous fresh

ZSM5/Fe 5% SSE fresh

ZSM5/Fe 13% Alkaline fresh

-100

100

300

500

700

900

1100

1300

1500

82 87 92 97 102 107

Inte

nsity

(a.u

.)

2 Theta (°)

H-ZSM5 freshZSM5/Fe 1% non-aqueous freshZSM5/Fe 5% SSE freshZSM5/Fe 13% alkaline fresh

0

2000

4000

6000

8000

10000

22 24 26

15

3.2 Catalytic performance of iron-ZSM5 catalysts

Catalytic measurements using H-ZSM5 as reference material resulted in selectivities of allyl

alcohol of approximately 1 %, as evident in Figure 5 which is in general agreement with other

reports of zeolite-catalysed glycerol conversion [4, 31, 38, 39]. The performance of the

synthesised Fe-ZSM5 catalysts was assessed on the basis of allyl alcohol production, disclosing

a relationship between iron loading and the activity of the catalyst towards allyl alcohol (Table

1 and Figure 5). The highest yield of allyl alcohol and the lowest average selectivity to acrolein

(Table 2) were obtained over the catalyst with the highest iron content (synthesised by alkaline

ion exchange) following six hours of time-on-stream catalytic measurements. Even though

experiments were conducted under conditions where complete conversion of glycerol was

achieved, conclusions can be drawn in terms of net selectivity of allyl alcohol over the catalysts

studied. While catalysts with an iron content ≤ 5 wt % afforded a low but virtually constant

allyl alcohol selectivity as function of time on stream, over the 13 wt % iron zeolite, allyl

alcohol is only detected in significant quantities following 4 hours of time on stream. This may

suggest that mass transfer limitations, changes in iron speciation with time on stream, poisoning

of acid sites due to coke deposition or/and that the reaction follows a different pathway when

iron is present in such high concentration.

16

Figure 5. Allyl alcohol selectivity as a function of time on stream over ZSM5/Fe catalysts. GHSV = 1240 h-1, Reactant: 35 wt % glycerol.

The decreased rate of formation of acrolein using the ZSM5/Fe 13 wt % catalyst could be

explained by a combination of two factors; one is the severity of the treatment conditions used

for iron deposition. This was reported by Possato et al. for their desilicated zeolites and was

attributed to changes in the density of Lewis and BrØnsted sites [31]. A second explanation is

also related to acid-base properties, as a decrease in acid site density was found to occur due to

iron deposition on the zeolite where the amount of iron exceeded the theoretical exchange

capacity of the material (Figure S1 ESI). Moreover, using the same catalyst, an increase in the

rate of formation of both acetic and propanoic acid was observed (Table 2). Even though H-

ZSM5 is a stable catalyst for glycerol conversion, iron deposition (where iron was present in

concentrations ≤ 5 wt %) reduces catalytic activity, as observed by a drop in glycerol

conversion after 6 hours of time on stream. However with the catalyst of higher iron content,

glycerol conversion remained constant at the end of the test (Table 2).

0

1

2

3

4

5

6

7

0 50 100 150 200 250 300 350 400

Ally

l alc

ohol

sele

ctiv

ity (%

)

Time on stream (min)

H-ZSM5ZSM5 1% non-aqueous IEH-ZSM5 5% SSEZSM5/Fe 13% alkaline IE

17

Table 2. Average product distribution over ZSM5/Fe catalysts and support during 6 hours of time on stream. GHSV = 1240 h-1, Temperature = 340 °C, Reactant: 35 wt % glycerol.

H-ZSM5 ZSM5/Fe 0.5 %

(non-aqueous IE)

ZSM5/Fe 5

% (SSE)

ZSM5/Fe 13 %

(alkaline IE)

Conversion (%) 98.9 96.8 84.8 99.9

Molar carbon

selectivity (%)

Allyl alcohol 0.7 1.4 2.7 2.7

Acrolein 13.1 11.9 8.2 7.0

Acrolein (gas phase) 28.9 22.9 20.1 11.2

Acetaldehyde 4.6 5.2 4.3 6.4

Acetaldehyde (gas

phase) 9.0 4.2 3.7 4.0

Hydroxyacetone 7.3 8.6 10.2 8.0

Acetic acid 0.7 0.6 1.1 2.1

Propanoic acid 0.3 0.4 0.7 3.1

Carbon dioxide 0.8 0.3 0.7 1.7

Carbon monoxide 0.2 0.4 0.4 1.0

Others 28.9a 38.9a 42.6b 47.0b

Coke 5.5 5.2 5.2 5.8

a: Others include 1,3-dioxan-5-ol and propanal 3-methoxy b: Others include 2,3 butanedione, glyceraldehyde, acetone and acetoin

3.3 Post reaction studies

Catalyst surface areas were significantly affected by the reaction, as evident in Figure 6 in

comparison with Figure 1. A reduction in surface area in the range of 92 - 94 % occurred with

18

low iron content catalysts, whereas the same parameter decreased by approximately 82 % over

the ZSM5 supported 13 wt % iron catalyst (Figure 6). Nevertheless, changes in surface area

were found to be reversible with catalyst regeneration (in air flow at 550 °C for five hours with

a heating rate of 1 °C·min-1) as observed for H-ZSM5 catalysts (Figure S2 ESI).

Figure 6. Nitrogen adsorption isotherms for used catalysts at -196 °C.

Post reaction analyses were also conducted using X-ray powder diffraction analysis, where a

change in the diffraction pattern was notable (Figure 7). Diffraction patterns of the catalysts

studied revealed that reflection doublets at 2θ = 23.08°, 23.15°; 26.49, 26.84; 35.77, 35.95 and

37.14, 37.51 were altered following the reaction, and appear as singlet reflections at 2θ = 23.09,

26.69, 35.86 and 37.45 respectively. In order to confirm that the structural change was due to

carbon deposition, experiments were conducted for up to six hours with water as feed and

nitrogen as carrier gas at 340 °C and atmospheric pressure. The XRD pattern of these samples

resembles those of fresh catalysts (Figure S3 ESI). Phase identification of the resultant patterns

showed an association between peak position and intensity of fresh catalysts to orthorhombic

crystals, while patterns of used catalysts corresponded to tetragonal crystal systems (not

0

10

20

30

40

50

60

0 0.2 0.4 0.6 0.8 1

Qua

ntity

ads

orbe

d (c

m3 /g

)

P/P0

H-ZSM5

ZSM5/Fe 1% non-aqueous IE

ZSM5/Fe 5% SSE

ZSM5/Fe 13% alkaline IE

19

shown). This observation is consistent with the work of Alvarez et al. in which cell parameter

calculations confirm the presence of carbonaceous deposits within the zeolite channels [40].

The phase change was proven to be reversible with reactivation in air at 550 °C for five hours

(Figure 7).

Figure 7. XRD patterns of used H-ZSM5 and ZSM5/Fe catalysts prepared by different methods.

In order to determine whether the alkaline ion exchange procedure influenced the nature of the

iron species deposited on ZSM5, a 70 wt % ZSM5/Fe catalyst was prepared using 2.5 g of H-

ZSM5 in an iron nitrate aqueous solution of 0.4 M ammonium hydroxide. X-ray diffraction

patterns of the resultant materials exhibited additional reflections (other than the associated to

MFI), disclosing the presence of the iron oxide phase hematite (Figure 8). This is in agreement

with previous reports in which iron hydroxides precipitates were observed to form hematite

following calcination [41].

0

5000

10000

15000

20000

25000

30000

35000

22 23 24 25 26 27 28

Inte

nsity

(a.u

.)

2 Theta (°)

H-ZSM5 freshH-ZSM5 usedH-ZSM5 used reactivated

0

2000

4000

6000

8000

10000

12000

14000

16000

22 24 26 28

Inte

nsity

(a.u

.)

2 Theta (°)

ZSM5/Fe 1% fresh

ZSM5/Fe 1% used

0

2000

4000

6000

8000

10000

12000

22 24 26 28

Inte

nsity

(a.u

.)

2 Theta (°)

ZSM5/Fe 5% freshZSM5/Fe 5% used

0

1000

2000

3000

4000

5000

6000

7000

22 23 24 25 26 27 28

Inte

nsity

(a.u

.)

2 Theta (°)

ZSM5/Fe 13% fresh

ZSM5/Fe 13% used

20

Figure 8. XRD patterns of H-ZSM5, Fe2O3a and 80 wt % Fe-Z-SM5 catalysts.

a Prepared by calcining Fe(NO3)3 in air flow at 400 °C for 4 hours.

3.4 Alkali metal deposition

The ZSM5 supported 13 % iron oxide catalyst was modified by the deposition of rubidium

nitrate, as basicity has been suggested to favour the formation of allyl alcohol when reacting

glycerol over iron catalysts [5, 7]. Despite the vast utilization of Cs exchanged zeolites, a Rb

salt was selected to conduct the ZSM5-supported iron catalysts modification based on previous

investigations with alumina-supported iron catalysts [5]. In this study, it was found that even

though ɣ-alumina/Fe/Cs resulted in the most stable catalyst, ɣ-alumina/Fe/Rb was highly

selective to allyl alcohol production from glycerol. Likewise, lower selectivity towards acrolein

was obtained over the latter catalyst in comparison to ɣ-alumina/Fe/Cs. At this stage the reason

for this catalytic behaviour is still unclear. In preliminary tests over H-ZSM5 catalysts,

deactivation was not an important issue following 360 min of catalytic measurement (Table 2)

which influenced in the selection of the Rb modified zeolite as opposed as the Cs treated

material.

0

2000

4000

6000

8000

10000

12000

14000

16000

5 15 25 35 45 55

Inte

nsity

(a.u

.)

2 Theta (°)

H-ZSM5 fresh

ZSM5/Fe 80% ion exchange (alkaline)

Iron oxide

21

As disclosed in Figure 9, the yield of allyl alcohol effectively doubled as result of the catalyst

treatment with rubidium. Simultaneously, a decrease in the yield of acrolein was observed with

the presence of the rubidium modifier.

Figure 9. a. Allyl alcohol and b. Acrolein yield as a function of time on stream over Rb modified Fe-ZSM5 catalysts. GHSV = 1240 h-1, Temperature: 340 °C, Reactant: 35 wt % glycerol.

Complete glycerol conversion was observed following 6 hours of time on stream. An apparent

reduction in coke formation was attributed to rubidium deposition. Over the 13 wt % iron oxide

on ZSM5 catalyst, the carbon content of the post reaction catalyst decreased from 11 wt %

0

2

4

6

8

10

12

0 100 200 300 400

Acr

olei

n yi

eld

(%)

Time on stream (min)

Fe-ZSM5

Fe-ZSM5 repeat

Fe-ZSM5/Rb

0

2

4

6

8

10

12

14

0 100 200 300 400

Ally

l alc

ohol

yie

ld (%

)

Time on stream (min)

Fe-ZSM5

Fe-ZSM5 repeat

Fe-ZSM5/Rb

a

b

22

(unmodified sample) to 10 wt % due to the alkaline metal modification. Changes in product

distribution are also thought to be a consequence of the reduction of acid sites, which promote

the formation of acrolein. If the concentration of acrolein is reduced, then the rate of side

reactions, such as the reaction between acrolein and glycerol or polymerisation of acrolein [4]

that lead to coke formation, are also reduced.

To study acid site density, ammonia was absorbed on both unmodified and modified iron ZSM5

catalysts, and the course of its desorption (TPD trace) is disclosed in Figure 10. Even though

the intensity of weak acid sites (150 - 300 °C) [42] remained unchanged following the alkaline

treatment of the iron catalyst, a significant reduction in the concentration of medium strength

acid sites (characterised by desorption of NH3 from the catalyst between 300 and 500 °C) [42]

was observed in the presence of rubidium.

Figure 10. Ammonia temperature program desorption profiles for ZSM5/Fe and ZSM5/Fe/Rb

catalysts, m/z = 16.

In spite of the significant influence of the rubidium modification on the rate of formation of

acrolein and allyl alcohol, little effect was observed on hydroxyacetone yields, particularly

0

20

40

60

80

100

120

140

0 100 200 300 400 500 600

Inte

nsity

(a.u

.)

Temperature (°C)

ZSM5/Fe(13%)/Rb

H-ZSM5/Fe(13%)

23

following 240 minutes of catalytic measurements (Figure 11). Previous studies on glycerol

conversion over unmodified and sodium-exchanged niobium oxide catalysts allowed

correlating acrolein yields with the concentration of Brønsted sites as well as hydroxyacetone

yields with the concentration of Lewis sites. The presence of alkali metals reduced Brønsted

acidity and therefore the rate of formation of acrolein, while the concentration of Lewis sites

remained constant so did hydroxyacetone yields [43]. Current ammonia TPD results confirmed

reduced acidity upon alkali addition. The study of Foo et al., other precedent literature [31, 44-

47] and current product distribution suggest that the rubidium modification of ZSM5-supported

iron catalysts mainly affects the concentration of Brønsted sites.

Figure 11. Hydroxyacetone yield as a function of time on stream over Rb modified Fe-ZSM5

catalysts. GHSV = 1240 h-1, Temperature: 340 °C, Reactant: 35 wt % glycerol.

Zeolite supported iron catalysts undergoing post synthesis modification resulted in stable

materials that supress acrolein formation and simultaneously enhance selectivity towards allyl

alcohol. For benchmarking purposes, Table 3 discloses glycerol conversion and selectivity

values obtained over different zeolites when converting glycerol operating under various

conditions. In studies conducted by other authors, molar carbon selectivity towards allyl

0

2

4

6

8

10

12

14

0 100 200 300 400

Hyd

roxy

acet

one

yiel

d (%

)

Time on stream (min)

Fe-ZSM5

Fe-ZSM5 repeat

Fe-ZSM5/Rb

24

alcohol does not exceed 2 % [4, 31, 38, 39]. Despite differences in catalysts and operating

conditions, hydroxyacetone selectivity values reported previously are in agreement with the

current results which also suggest that the iron exchange has considerably influence on

Brønsted sites as opposed to Lewis sites.

Table 3. Product distribution obtained over zeolites for glycerol conversion

Catalyst H-ZSM5 b HZSM5 c HZSM5 c SAPO-11 HZSM5c/Fe/Rb

Conversion (%) 57.0 36.3 74.0 88.0 99.9

Allyl alcohol

Selectivity (mol %) 2.0 0.9 1.1 1.5 11.9

Acrolein selectivity

(mol %) 67.0 45.8 22.0 62.0 3.8

Hydroxyacetone

selectivity (mol %) 12.0 11.0 4.0 10.0 9.0

Temperature 320 315 300 280 340

GHSVa 1438 h-1 1392 h-1 1177 h-1 43 h-1 110 h-1

Reference [39] [37] [31] [4] This work

a: GHSV calculated respect to glycerol b: SiO2/Al2O3 = 30 c: SiO2/Al2O3 = 23

Iron speciation has been observed by Groen et al. for iron-containing zeolites undergoing

desilication through post synthesis alkaline treatment [48]. Figure 12.a shows the XPS survey

spectrum of the ZSM5/Fe 1% catalyst as prepared, in which iron was not detected. For the

ZSM5/Fe 13% and ZSM5/Fe/Rb catalysts, Fe3+ was found to be present as evident by the

Fe2p3/2 peak at 711.5 eV, the Fe2p3/2 satellite peak at 719.3 eV, the Fe2p1/2 peak at 725.0 eV

and the Fe2p1/2 satellite peak at 733.7 eV. In the ZSM5/Fe 5% catalyst, the presence of Fe2+

25

and Fe3+ was confirmed by the Fe2p3/2 peak at 711.0 eV, the Fe2p1/2 peak at 724.4 eV and the

absence of satellite peaks as disclosed in Figure 12.b. [49-51]. The peak at 533.1 eV, present

in all cases, has been attributed as reported elsewhere [52, 53] to ZSM5 lattice oxygen. The

shoulder at lower binding energy values (530.6 eV) observable for high iron content catalysts

suggests the presence of extra framework iron oxide species which is in line with previous

works [52, 53].

Figure 12 a. XPS survey spectrum of ZSM5/Fe 1% catalyst. b. High resolution O1s spectra of

ZSM5 supported iron catalysts. c. High resolution Fe2p spectra of ZSM5 supported iron

catalysts.

0.0E+00

2.0E+05

4.0E+05

6.0E+05

8.0E+05

1.0E+06

1.2E+06

05001000

Inte

nsity

(a.

u.)

Binding energy (eV)

O1s

N1s C1s Al2p

0.00E+00

4.00E+04

8.00E+04

1.20E+05

1.60E+05

2.00E+05

520525530535540545

Inte

nsity

(a.

u.)

Binding energy (eV)

ZSM5/Fe 1%ZSM5/Fe 5%ZSM5/Fe 13% freshZSM5/Fe 13% usedZSM5/Fe/Rb 13%

10000

15000

20000

25000

30000

35000

40000

700710720730740

Inte

nsity

(a.u

.)

Binding energy (eV)

ZSM5/Fe 5%

ZSM5/Fe 13% fresh

ZSM5/Fe 13% used

ZSM5/Fe/Rb 13%

Si2p

Fe2p3/2

Satellite

Fe2p1/2

a b

c

26

UV-vis spectroscopy allowed distinguishing between different Fe3+ species present in the

studied catalysts. As evident on Figure 13.a, the Rb modification results in an increase in

isolated tetrahedrally coordinated Fe3+ species (associated with the oxygen to iron charge

transfer band at 240 nm) as well as of Fe3+ species coordinated with high number of oxygen

ligands (band at 290 nm) at the expense of oligomeric iron clusters (bands between 300 nm

and 400 nm) and iron particles (bands above 400 nm) [48, 54]. In agreement with the work of

Yue et al. with iron modified ZSM5 catalysts, the absence of reflections corresponding to iron

oxides was thought to be due to the size of such particles which do not exceed 4 nm [52]. Based

on product distribution these isolated Fe3+ species were thought to enhance the rate of formation

of allyl alcohol from glycerol.

Figure 13 a. UV-vis spectra of ZSM5/Fe and ZSM5/Fe/Rb fresh catalysts. b. UV-vis spectra

of ZSM5/Fe used catalyst.

Following catalytic measurements over ZSM5/Fe 13%, the used catalyst was analysed by XPS.

XPS spectrum suggests iron ions in divalent oxidation state, as evident by the Fe2p3/2 peak at

711.0 eV, the Fe2p1/2 peak at 724.4 eV and the absence of satellite peaks (Figure 12.b).

Likewise, the UV-vis spectrum of the used catalyst is comparable to that of Fe3O4 [55] (Figure

13.b) and therefore is in agreement with the current XPS results. This is evidence of changes

in iron speciation with time on stream.

0

5

10

15

20

25

200 400 600 800 1000

F (R

)

λ (nm)

ZSM5/Fe

ZSM5/Fe/Rb

0

10

20

30

40

50

60

200 400 600 800 1000

F (R

)

λ (nm)

ZSM5/Fe 13% used

a b

27

3.5 Kinetic diameter calculations

Studies conducted by Gu et al. [56], over several zeolites with differing channel configurations

and diameters, suggest that, optimal catalysts for glycerol conversion to acrolein were those

with channel dimensions slightly larger than the molecular diameter of glycerol (5.0 Å), as for

instance H-ZSM5 (straight channels with dimensions of 5.1 Å × 5.5 Å). Their observation was

based on the existence of straight channels in H-ZSM5, which facilitate enhanced rates of

diffusion when compared to more complex pore architectures. Moreover, H-ZSM5 steric

hindrance (found to play a more important role than acidity) limits coke formation. The same

kinetic (4 Å) diameter has been reported for other alcohols and their corresponding aldehydes

such as ethanol and acetaldehyde [57, 58]. However, to the best of our knowledge, the kinetic

diameters for allyl alcohol and acrolein have not been reported in the literature. Kinetic

diameters for biomass feedstocks and their possible products were calculated by Jae et al. [2]

using equations by Bird et al. [59]. Adopting a similar procedure, we have estimated the

diameters for the desired product and primary by-product of the reaction of glycerol over

zeolites. As found previously [2], there are variations between the estimation based on critical

volume, critical temperature-pressure and molecular weight respectively (

28

Table 4). However regardless the method employed in the calculations, similar estimates of the

molecular dimensions of acrolein and allyl alcohol products were determined.

29

Table 4. Acrolein and allyl alcohol kinetic diameters calculated according to Jae et al. [2].

Allyl alcohol Acrolein Units and references

퐾 = 0.841푉 5.9 5.7 Critical volume in cm3·mol-1

퐾 = 2.44푇푃

5.2 5.2 Critical temperature in K

Critical Pressure in atm

퐾 = 1.234푀 4.8 4.7 -

Acrolein and allyl alcohol appear to have similar transport characteristics controlling their rate

of diffusion in zeolite channels, given by minor differences in their kinetic diameters. However,

channel structure and dimensions influence coke formation [56]. Increased mesoporosity was

observed with the zeolite with the highest iron content, which correlates with an increased rate

of coke formation and reduced rate of acrolein formation. This is in agreement with the work

of other authors [56] in which selectivity to acrolein decreased over catalysts of larger

diameters pore channels. This behaviour was attributed to secondary reactions (some leading

to coke production) and reduced interaction of the central OH in the glycerol molecule with

the bridging OH on the catalyst compared to smaller channel diameter zeolites. For instance,

over materials such as ferrierite, coke production increased and the rate of formation of acrolein

decreased compared to results obtained over H-ZSM5 at the same reaction conditions [38].

Moreover materials such as MCM-41 are well known to catalyse the formation of polyglycerols

[3, 60]. The production of allyl alcohol over this catalyst does not depend strongly on transport

limitations, but seems to be mainly governed by acidity (as evident in Figure S1 in the ESI by

ammonia temperature desorption on H-ZSM5 and H-ZSM5/Fe catalysts) and/or redox site

density.

30

A series of experiments were conducted to study the effect of transport limitations on the

reaction, initially by reducing the catalyst weight by 90 % in order to operate under low

conversion conditions.

A preliminary test was used as base line in the detection of internal and external mass transfer

limitations. This test involved using 500 mg of ZSM5/Fe 13 % catalyst, feeding a 35 wt %

glycerol solution at a flow rate of 0.75 mL min-1, with nitrogen as gas carrier at a flow rate of

180 mL min-1 for a GHSV of 19250 h-1. In a separate experiment, 500 mg of ZSM5/Fe 13 %

catalyst was sized between 600 µm to 1 mm, while the remainder operating conditions

remained unchanged. With a larger particle size, conversion is expected to drop if the reaction

is limited by internal diffusion [61]. Following 360 min of catalytic measurements at these

conditions, glycerol conversion was approximately 25.0 % as shown in

31

Table 5, which is comparable to tests where catalysts with smaller particle size were used,

suggesting no internal diffusion limitations.

Moreover, a follow up series of experiments involved increasing the catalyst mass up to 1 g,

where the flow rates of glycerol solution and gas carrier were adjusted such that the GHSV was

maintained constant. Under these conditions, an increase in conversion is expected if the

reaction is limited by external diffusion [61]. As shown in

32

Table 5, glycerol conversion levels roughly double over 1 g of catalyst with respect to tests in

which 500 mg of catalyst was used, suggesting the reaction was under external diffusion control

conditions.

33

Table 5. Glycerol conversion as a function of time on stream over ZSM5/Fe 13 %. GHSV =

19250 h-1, Temperature: 340 °C, Reactant: 35 wt % glycerol.

Time (min)

Conversion (%)

500 mg (250 μm - 600

μm)

500 mg (600 μm -

1mm)

1 g (250 μm - 600

μm)

60 26.6 26.2 50.6

120 28.0 24.1 51.4

180 23.8 27.5 50.5

240 26.0 25.8 54.8

300 25.7 26.1 54.0

360 25.8 25.0 54.0

In the light of these results, additional experiments were conducted in order to determine

whether the induction period for allyl alcohol production observed in Figure 5 was exclusively

a consequence of external mass transfer limitations or if it was also influenced by intrinsic

kinetics. These tests were designed to operate in a differential reactor mode which often

simplify kinetic investigations and where transport and heat limitations are negligible [62]. As

shown in Table 6, similarities were found in glycerol conversion values when 50 mg and 100

mg of ZSM5/Fe 13 % catalyst are used. Interestingly, Table 7, an induction period for allyl

alcohol selectivity is observed. These suggest that the formation of ally alcohol at certain time

on stream is not only a result of external transfer limitations (

34

Table 5), but also a consequence of intrinsic kinetics (Table 7).

Table 6. Glycerol conversion as a function of time on stream over ZSM5/Fe 13 %. GHSV =

192500 h-1, Temperature: 340 °C, Reactant: 35 wt % glycerol.

Time (min)

Conversion (%)

50 mg (250 μm - 600 μm) 100 mg (250 μm - 600 μm)

60 5.5 5.1

120 5.7 5.0

180 2.6 3.1

240 1.8 2.1

300 1.7 1.4

Table 7. Allyl alcohol selectivity as a function of time on stream over ZSM5/Fe 13 %. GHSV

= 192500 h-1, Temperature: 340 °C, Reactant: 35 wt % glycerol.

Time (min)

Selectivity (%)

50 mg (250 μm - 600 μm) 100 mg (250 μm - 600 μm)

60 24.3 23.5

120 14.4 22.8

180 41.8 29.6

240 61.1 50.1

35

300 61.8 60.1

Based on these results and on product distribution over different catalysts as function of time

on stream, the induction period required for the production of allyl alcohol it is not exclusively

attributed to transfer mass limitations, but also two events might have an effect: the reduction

to Fe2+ and the poisoning of acid sites on the catalysts by means of coke deposition. Evidence

for the former is the shorter induction period for allyl alcohol formation when ZSM5/Fe 5%

catalysts prepared by SSE catalyse the reaction. Fe2+ was identified to be present in this catalyst

by XPS (Figure 12.b). Evidence for the latter are ammonia TPD of used ZSM5/Fe 13 wt %,

which suggests a substantial reduction of both weak and medium acid sites (Figure S4 ESI)

and the shorter induction time required for allyl alcohol production when glycerol conversion

is conducted over ZSM5/Fe/Rb 13% catalysts, in which the concentration of acid sites was

found to be reduced by the presence of the alkali metal.

3.6 Insights into the reaction mechanism for allyl alcohol formation

When reacting glycerol using alumina as catalyst support instead of H-ZSM5 and operating at

similar conditions, a reduction on the rate of formation of hydroxyacetone coincided with an

increase in allyl alcohol yields. The alkali metal modification of alumina supported catalysts

increased the production of carbon monoxide and hydrogen from glycerol which were thought

to act as sacrificial reductants resulting in the formation of allyl alcohol and carbon dioxide at

the expense of hydroxyacetone [5]. Even though in the current experiments (using zeolites as

support), the use of a rubidium-modified catalyst increased the yield of carbon monoxide (from

0.9 % to 1.5 %) and carbon dioxide (from 1.7 % to 2.3 %), selectivity towards hydroxyacetone

remained virtually constant suggesting a different mechanism.

36

Iron oxide has been reported to catalyse hydrogen transfer reactions in the presence of glycerol

[20], as well as oxidation reactions of glycerol dehydration products [63]. Moreover, iron has

been found to chelate with a diol resulting in the formation of 1,3-propanediol from glycerol

[64]. Previous studies proposed 1,3-propanediol as a possible intermediate in the synthesis of

allyl alcohol from glycerol [65]. Additionally, ɣ-Fe2O3 has been reported to catalyse

deoxygenation reactions via disproportionation [66].

In a recent review of the available literature in polyols transformations, the van Krevelen

diagram was used to point out three pathways for allyl alcohol formation from glycerol. A first

possibility is the single dehydration of the glycerol molecule to hydroxyacetone and its

subsequent deoxygenation. However, in this reaction, acrolein is typically produced in high

selectivities. The second route is glycerol deoxygenation resulting in the formation of 1,2-

propanediol which is then dehydrated to allyl alcohol. The deoxydehydration (DODH)

reaction, generally catalysed by rhenium complexes, arises in an attempt to overcome the

difficulties in controlling selectivity in the other reaction pathways [67].

In the current experiments, when H-ZSM5 catalysed the conversion of glycerol, high

selectivity towards 1,3-dioxan-5-ol was observed. This latter compound is most likely to be

generated by the condensation of a glycerol molecule with a formaldehyde molecule [68, 69].

For the aldehyde to be present in the system in significant quantities to produce 1,3-dioxan-5-

ol, the 1,3-dehydration of glycerol must occur [70]. Likewise acrolein was produced in high

yields over this catalyst. However, when zeolite supported iron materials catalyse the same

reaction, the rate of formation of 1,3-dioxan-5-ol was reduced and selectivity towards

compounds such as acetic acid, propanoic acid, acetone, glyceraldehyde, 1,3-propanediol and

1,3-dioxolane-4-methanol,2-ethyl (product of condensation of 1,3-propanediol and acrolein

[65]) increased, which suggests redox processes. This coincided with an enhancement in the

rate of formation of allyl alcohol. Based on product distribution, a two-step reaction mechanism

37

was thought to be involved in the formation of allyl alcohol from glycerol over iron exchanged

zeolites. Firstly, the glycerol molecule undergoes single 1,2-dehydration for which acid sites

on the zeolite are required. Excessive 1,2-dehydration results in the formation of acrolein while

1,3-dehydration allows 1,3-dioxan-5-ol to be formed. The second step is a deoxygenation

process where iron provides redox sites for the reaction to take place. Evidence of this are the

decreased rates of formation of acrolein and 1,3-dioxan-5-ol in the presence of iron and the

formation of the oxygenates mentioned above. For the formation of these compounds, reactions

that involve the removal of oxygen atoms from glycerol dehydration products are likely to

occur. Such reactions were thought to result in the formation of allyl alcohol.

In summary the role and function of iron in the formation of allyl alcohol is to provide redox

sites for deoxygenation and hydrogen transfer reactions to occur. Rubidium, instead, poisons

acid sites in the catalyst which prevents excessively dehydration to take place. Even though

dehydration is necessary for the formation of allyl alcohol, the double glycerol dehydration

leads to the formation of acrolein as opposed to allyl alcohol.

4. Conclusions

Using an alkaline solution instead of methanolic to conduct ion exchange in HZSM-5 resulted

in the synthesis of a 13 wt % iron zeolite, which performed better for allyl alcohol production

compared to catalysts synthesised by other methods. The severity of the preparation procedure

promoted partial dissolution of both framework silicon and aluminium, affecting the micropore

volume and engendered the zeolite with mesoporosity. Post synthesis modification by rubidium

deposition resulted in additional enhancement in the yield of allyl alcohol and decreased the

yield of acrolein. Reduction in the rate and extent of coke formation were attributable to the

rubidium deposition, which is speculated to be due to a decrease in the concentration of medium

acid sites, which was confirmed by ammonia TPD. Kinetic diameter calculations showed that

38

the production of allyl alcohol over the assessed catalysts does not depend strongly on pore

size distribution, but seems to be mainly governed by acidity and/or redox site density.

5. Acknowledgements

Funding by African Explosives Ltd. is gratefully acknowledged. GS would like to thank the

University of Newcastle for a post graduate research scholarship. Authors thank the Electron

Microscope and X-ray Unit of the University of Newcastle for the use of their facilities.

Authors thank Ms. Jane Hamson for her assistance with ICP analyses and Mr. Dylan Cuskelly

for his assistance with XRD measurements.

6. References

[1] M.d.S. Machado, J. Perez-Pariente, E. Sastre, D. Cardoso, A. De Guerenu, Selective

synthesis of glycerol monolaurate with zeolitic molecular sieves, Appl. Catal., A, 203 (2000)

321-328.

[2] J. Jae, G.A. Tompsett, A.J. Foster, K.D. Hammond, S.M. Auerbach, R.F. Lobo, G.W.

Huber, Investigation into the shape selectivity of zeolite catalysts for biomass conversion, J.

Catal., 279 (2011) 257-268.

[3] J.M. Clacens, Y. Pouilloux, J. Barrault, Selective etherification of glycerol to polyglycerols

over impregnated basic MCM-41 type mesoporous catalysts, Appl. Catal., A, 227 (2002) 181-

190.

[4] W. Suprun, M. Lutecki, T. Haber, H. Papp, Acidic catalysts for the dehydration of glycerol:

Activity and deactivation, J. Mol. Catal. A: Chem., 309 (2009) 71-78.

39

[5] G. Sánchez, J. Friggieri, C. Keast, M. Drewery, B.Z. Dlugogorski, E. Kennedy, M.

Stockenhuber, The effect of catalyst modification on the conversion of glycerol to allyl alcohol,

Appl. Catal. B, 152–153 (2014) 117-128.

[6] T. Yoshikawa, T. Tago, A. Nakamura, A. Konaka, M. Mukaida, T. Masuda, Investigation

of reaction routes for direct conversion of glycerol over zirconia–iron oxide catalyst, Res Chem

Intermed, 37 (2011) 1247-1256.

[7] A. Konaka, T. Tago, T. Yoshikawa, A. Nakamura, T. Masuda, Conversion of glycerol into

allyl alcohol over potassium-supported zirconia–iron oxide catalyst, Appl. Catal., B, 146

(2014) 267-273.

[8] J.C. Groen, J.A. Moulijn, J. Pérez-Ramírez, Decoupling mesoporosity formation and

acidity modification in ZSM-5 zeolites by sequential desilication–dealumination, Micropor.

Mesopor. Mat., 87 (2005) 153-161.

[9] L. Kubelková, S. Beran, A. Malecka, V.M. Mastikhin, Acidity of modified Y zeolites:

Effect of nonskeletal Al, formed by hydrothermal treatment, dealumination with SiCl4, and

cationic exchange with Al, Zeolites, 9 (1989) 12-17.

[10] Y. Sun, S.M. Campbell, J.H. Lunsford, G.E. Lewis, D. Palke, L.M. Tau, The catalytic

conversion of methyl chloride to ethylene and propylene over phosphorus-modified Mg-ZSM-

5 zeolites, J. Catal., 143 (1993) 32-44.

[11] M. Ogura, S.-y. Shinomiya, J. Tateno, Y. Nara, M. Nomura, E. Kikuchi, M. Matsukata,

Alkali-treatment technique — new method for modification of structural and acid-catalytic

properties of ZSM-5 zeolites, Appl. Catal., A, 219 (2001) 33-43.

[12] K.S. Walton, M.B. Abney, M. Douglas LeVan, CO2 adsorption in Y and X zeolites

modified by alkali metal cation exchange, Micropor. Mesopor. Mat., 91 (2006) 78-84.

[13] NPCS, B.o. Consultants, Engineers, Industrial alcohol technology handbook, Asia Pacific

Business Press Inc.2010.

40

[14] Lyondell, Data review and assessment for allyl alcohol, 2005.

[15] Lyondell, Application data allyl alcohol, 2006.

[16] D.T. Johnson, K.A. Taconi, The glycerin glut: Options for the value-added conversion of

crude glycerol resulting from biodiesel production, Environ. Prog., 26 (2007) 338-348.

[17] J.G. Speight, Chemical and process design handbook, (2002).

[18] N. ChemSystems, Evolving propylene sources solution to supply shortages?, 2012.

[19] J. Caner, Z. Liu, Y. Takada, A. Kudo, H. Naka, S. Saito, Synthesis of propylene from

renewable allyl alcohol by photocatalytic transfer hydrogenolysis, Catal. Sci. Tech., 4 (2014)

4093-4098.

[20] Y. Liu, H. Tuysuz, C.-J. Jia, M. Schwickardi, R. Rinaldi, A.-H. Lu, W. Schmidt, F. Schuth,

From glycerol to allyl alcohol: iron oxide catalyzed dehydration and consecutive hydrogen

transfer, Chem. Commun., 46 (2010) 1238-1240.

[21] Ü. Arda, Conversion of glycerol to the valuable intermediates acrolein and allyl alcohol

in the presence of heterogeneous catalysts, RWTH Aachen University, 2009, pp. 137.

[22] J.A.Z. Pieterse, S. Booneveld, R.W. van den Brink, Evaluation of Fe-zeolite catalysts

prepared by different methods for the decomposition of N2O, Appl. Catal., B, 51 (2004) 215-

228.

[23] J.P.H. Li, E. Kennedy, M. Stockenhuber, Oxidative coupling and hydroxylation of phenol

over transition metal and acidic zeolites: Insights into catalyst function, Catal. Lett., 144 (2014)

9-15.

[24] M. Ghoorah, B.Z. Dlugogorski, H.C. Oskierski, E.M. Kennedy, Study of thermally

conditioned and weak acid-treated serpentinites for mineralisation of carbon dioxide, Miner.

Eng., 59 (2014) 17-30.

41

[25] M.L.M. Bonati, R.W. Joyner, M. Stockenhuber, A temperature programmed desorption

study of the interaction of acetic anhydride with zeolite beta (BEA), Catal. Today, 81 (2003)

653-658.

[26] G. Sánchez, J. Friggieri, A.A. Adesina, B.Z. Dlugogorski, E.M. Kennedy, M.

Stockenhuber, Catalytic conversion of glycerol to allyl alcohol; effect of a sacrificial reductant

on the product yield, Catal. Sci. Tech., (2014).

[27] M. Iwamoto, H. Yahiro, Y. Torikai, T. Yoshioka, N. Mizuno, Novel preparation method

of highly copper ion-exchanged ZSM-5 zeolites and their catalytic activities for NO

decomposition, Chem. Lett., (1990) 1967-1970.

[28] L.T. Tsikoza, E.V. Matus, Z.R. Ismagilov, V.A. Sazonov, V.V. Kuznetsov, Correlation

between the Cu/Al value and activity of Cu-ZSM-5 catalysts and the chemical nature of the

starting copper salt, Kinet. Catal., 46 (2005) 613-617.

[29] M. Iwamoto, H. Yahiro, Y. Mine, S. Kagawa, Excessively copper ion-exchanged ZSM-5

zeolites as highly active catalysts for direct decomposition of nitrogen monoxide, Chem. Lett.,

(1989) 213-216.

[30] M. Schreier, S. Teren, L. Belcher, J.R. Regalbuto, J.T. Miller, The nature of

‘overexchanged’ copper and platinum on zeolites, Nanotechnology, 16 (2005) 582-591.

[31] L.G. Possato, R.N. Diniz, T. Garetto, S.H. Pulcinelli, C.V. Santilli, L. Martins, A

comparative study of glycerol dehydration catalyzed by micro/mesoporous MFI zeolites, J.

Catal., 300 (2013) 102-112.

[32] I. Melian-Cabrera, S. Espinosa, J. Groen, F. Kapteijn, J. Moulijn, Utilizing full-exchange

capacity of zeolites by alkaline leaching: Preparation of Fe-ZSM5 and application in N2O

decomposition, J. Catal., 238 (2006) 250-259.

[33] C. Weidenthaler, W. Schmidt, Thermal stability and thermal transformations of Co2+-or

Ni2+-exchanged zeolites A, X, and Y, Chem. Mat., 12 (2000) 3811-3820.

42

[34] C. Resini, T. Montanari, L. Nappi, G. Bagnasco, M. Turco, G. Busca, F. Bregani, M.

Notaro, G. Rocchini, Selective catalytic reduction of NOx by methane over Co-H-MFI and Co-

H-FER zeolite catalysts: characterisation and catalytic activity, J. Catal., 214 (2003) 179-190.

[35] G. Ertl, H. Knozinger, J. Weitkamp, Handbook of heterogeneous catalysis, Wiley Online

Library, 1997.

[36] J. Perez-Ramirez, C.H. Christensen, K. Egeblad, C.H. Christensen, J.C. Groen,

Hierarchical zeolites: enhanced utilisation of microporous crystals in catalysis by advances in

materials design, Chem. Soc. Rev., 37 (2008) 2530-2542.

[37] E. Urquieta-González, L. Martins, R. Peguin, M. Batista, Identification of extra-

framework species on Fe/ZSM-5 and Cu/ZSM-5 catalysts typical microporous molecular

sieves with zeolitic structure, Mater. Res., 5 (2002) 321-327.

[38] Y.T. Kim, K.-D. Jung, E.D. Park, A comparative study for gas-phase dehydration of

glycerol over H-zeolites, Appl. Catal., A, 393 (2011) 275-287.

[39] C.-J. Jia, Y. Liu, W. Schmidt, A.-H. Lu, F. Schüth, Small-sized HZSM-5 zeolite as highly

active catalyst for gas phase dehydration of glycerol to acrolein, J. Catal., 269 (2010) 71-79.

[40] A.G. Alvarez, H. Viturro, R.D. Bonetto, Structural changes on deactivation of ZSM-5. A

study by X-ray powder diffraction, Mater. Chem. Phys., 32 (1992) 135-140.

[41] S. Sahoo, K. Agarwal, A. Singh, B. Polke, K. Raha, Characterization of γ-and α-Fe2O3

nano powders synthesized by emulsion precipitation-calcination route and rheological

behaviour of α-Fe2O3, Int. J. Eng., Sci. Tech., 2 (2010).

[42] M.H. Haider, N.F. Dummer, D. Zhang, P. Miedziak, T.E. Davies, S.H. Taylor, D.J.

Willock, D.W. Knight, D. Chadwick, G.J. Hutchings, Rubidium- and caesium-doped

silicotungstic acid catalysts supported on alumina for the catalytic dehydration of glycerol to

acrolein, J. Catal., 286 (2012) 206-213.

43

[43] G.S. Foo, D. Wei, D.S. Sholl, C. Sievers, Role of Lewis and Brønsted acid sites in the

dehydration of glycerol over niobia, ACS Catalysis, 4 (2014) 3180-3192.

[44] A. Alhanash, E.F. Kozhevnikova, I.V. Kozhevnikov, Gas-phase dehydration of glycerol

to acrolein catalysed by caesium heteropoly salt, Appl. Catal. A, 378 (2010) 11-18.

[45] Z. Wu, Y. Mao, M. Song, X. Yin, M. Zhang, Cu/boehmite: A highly active catalyst for

hydrogenolysis of glycerol to 1,2-propanediol, Catal. Comm., 32 (2013) 52-57.

[46] Y.T. Kim, K.-D. Jung, E.D. Park, Gas-phase dehydration of glycerol over silica–alumina

catalysts, Appl. Catal., B, 107 (2011) 177-187.

[47] P. Lauriol-Garbey, G. Postole, S. Loridant, A. Auroux, V. Belliere-Baca, P. Rey, J.M.M.

Millet, Acid–base properties of niobium-zirconium mixed oxide catalysts for glycerol

dehydration by calorimetric and catalytic investigation, Appl. Catal., B, 106 (2011) 94-102.

[48] J.C. Groen, L. Maldonado, E. Berrier, A. Brückner, J.A. Moulijn, J. Pérez-Ramírez,

Alkaline treatment of iron-containing MFI zeolites. Influence on mesoporosity development

and iron speciation, J. Phys. Chem. B, 110 (2006) 20369-20378.

[49] M. Descostes, F. Mercier, N. Thromat, C. Beaucaire, M. Gautier-Soyer, Use of XPS in

the determination of chemical environment and oxidation state of iron and sulfur samples:

constitution of a data basis in binding energies for Fe and S reference compounds and

applications to the evidence of surface species of an oxidized pyrite in a carbonate medium,

Appl. Surf. Sci., 165 (2000) 288-302.

[50] N.S. McIntyre, D.G. Zetaruk, X-ray photoelectron spectroscopic studies of iron oxides,

Anal. Chem., 49 (1977) 1521-1529.

[51] A. Grosvenor, B. Kobe, M. Biesinger, N. McIntyre, Investigation of multiplet splitting of

Fe 2p XPS spectra and bonding in iron compounds, Surf. Interface Anal., 36 (2004) 1564-1574.

44

[52] Y. Yue, H. Liu, P. Yuan, C. Yu, X. Bao, One-pot synthesis of hierarchical FeZSM-5

zeolites from natural aluminosilicates for selective catalytic reduction of NO by NH3, Sci. Rep.,

5 (2015).

[53] R.B. Borade, A. Adnot, S. Kaliaguine, Acid sites in Al-ZSM-22 and Fe-ZSM-22, Zeolites,

11 (1991) 710-719.

[54] M. Schwidder, M.S. Kumar, A. Brückner, W. Grünert, Active sites for NO reduction over

Fe-ZSM-5 catalysts, Chem. comm., (2005) 805-807.

[55] H. Zhu, E. Zhu, G. Ou, L. Gao, J. Chen, Fe3O4–Au and Fe2O3–Au hybrid nanorods: Layer-

by-layer assembly synthesis and their magnetic and optical properties, Nanoscale Res. Lett., 5

(2010) 1755-1761.

[56] Y. Gu, N. Cui, Q. Yu, C. Li, Q. Cui, Study on the influence of channel structure properties

in the dehydration of glycerol to acrolein over H-zeolite catalysts, Appl. Catal., A, 429–430

(2012) 9-16.

[57] H. Wu, Q. Gong, D.H. Olson, J. Li, Commensurate adsorption of hydrocarbons and

alcohols in microporous metal organic frameworks, Chem. Rev., 112 (2012) 836-868.

[58] I. Halasz, M. Agarwal, B. Marcus, W.E. Cormier, Molecular spectra and polarity sieving

of aluminum deficient hydrophobic H-Y zeolites, Micropor. Mesopor. Mat., 84 (2005) 318-

331.

[59] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport phenomena, John Wiley & Sons2007.

[60] A. Behr, J. Eilting, K. Irawadi, J. Leschinski, F. Lindner, Improved utilisation of

renewable resources: new important derivatives of glycerol, Green Chem., 10 (2008) 13-30.

[61] J.-F. Le Page, J. Cosyns, P. Courty, Applied heterogeneous catalysis: design, manufacture,

use of solid catalysts, Editions Technip1987.

[62] M.A. Vannice, W.H. Joyce, Kinetics of catalytic reactions, Springer2005.

45

[63] F. Wang, J. Xu, J.L. Dubois, W. Ueda, Catalytic oxidative dehydration of glycerol over a

catalyst with iron oxide domains embedded in an iron orthovanadate phase, ChemSusChem, 3

(2010) 1383-1389.

[64] J. Chaminand, L. aurent Djakovitch, P. Gallezot, P. Marion, C. Pinel, C. Rosier, Glycerol

hydrogenolysis on heterogeneous catalysts, Green Chem., 6 (2004) 359-361.

[65] L. Huang, Y. Zhu, H. Zheng, G. Ding, Y. Li, Direct conversion of glycerol into 1, 3-

propanediol over Cu-H4SiW12O40/SiO2 in vapor phase, Catal. Lett., 131 (2009) 312-320.

[66] E.J. Grootendorst, Y. Verbeek, V. Ponec, The role of the Mars and Van Krevelen

mechanism in the selective oxidation of nitrosobenzene and the deoxygenation of nitrobenzene

on oxidic catalysts, J. Catal., 157 (1995) 706-712.

[67] J.R. Dethlefsen, P. Fristrup, Rhenium-catalyzed deoxydehydration of diols and polyols,

ChemSusChem, 8 (2015) 767-775.

[68] J. Deutsch, A. Martin, H. Lieske, Investigations on heterogeneously catalysed

condensations of glycerol to cyclic acetals, J. Catal., 245 (2007) 428-435.

[69] J. Deleplanque, J.-L. Dubois, J.-F. Devaux, W. Ueda, Production of acrolein and acrylic

acid through dehydration and oxydehydration of glycerol with mixed oxide catalysts, Catal.

Today, 157 (2010) 351-358.

[70] M.R. Nimlos, S.J. Blanksby, X. Qian, M.E. Himmel, D.K. Johnson, Mechanisms of

glycerol dehydration, J. Phys. Chem. A, 110 (2006) 6145-6156.